
AegisOps is an agentic AI copilot for infrastructure and operations teams that combines anomaly detection, autonomous investigation, and mechanical validation in one auditable pipeline. It generates synthetic service telemetry, evaluates three detection approaches—Naive z-score/EWMA, Isolation Forest, and LSTM Autoencoder—and scores them across obvious spikes, gradual drift, and subtle correlated anomalies. Flagged incidents are investigated by a Claude tool-use agent that queries metrics, logs, and similar incidents before producing a structured incident report. A mechanical validator then checks every evidence claim against the tools actually called, requiring real citations and concrete values. The system exposes the same tool implementations through REST APIs and provides a React dashboard for incidents, traces, validation warnings, and tiered evaluation results.
16 Sep 2026